Want to see more threads and images? Ask Bernd!
Bernd Thu, 27 Feb 2025 17:06:04 GMT No. 25524067 [Kohl] [Report thread]
IMG_20250227_133010_149.png
6.53 kB, 479x431
1512368223001.jpg
27.54 kB, 487x423
Does bernd know anything about digital signal filtering? I have some professional signal-processing software, it has all tools for filtering and analysis, and I have a simple case of noise. At least I think it is simple, since I can see the repeating pattern, and I know what is signal and what is not. What I don't have is the education to design a filter which would work. I messed around with the tool, it has a dozen different filters (low pass, two band, three band, three band with a notch, infinite impulse response, some named filters), and each has many settings. I think I need a rather low frequency filter (50hz?), since the source is likely power supply, but the tutorial I tried for that case just gave me garbage, and didn't explain much.
Total posts: 122, files: 27 (Drowned at Fri, 28 Feb 2025 19:49:22 GMT)
Bernd Thu, 27 Feb 2025 17:09:15 GMT No. 25524085 >>25524130
>>25524067 What are you planning on using the signal for exactly? How precise does it need to be?
Bernd Thu, 27 Feb 2025 17:16:41 GMT No. 25524126 >>25524170
>>25524067 >source is likely power supply Get a better power supply?
Bernd Thu, 27 Feb 2025 17:16:59 GMT No. 25524127 >>25524166
just make a 180 phase-shifted copy the sum it with the original nah?
Bernd Thu, 27 Feb 2025 17:17:14 GMT No. 25524130 >>25524160 >>25525505
>>25524085 This specific recording is an example of noise, my signal here is in the amplitude and auc of the peak. This one will work without filtering. But I also have other recordings where I have random downward peaks much like those ones but about 10 times smaller.
Bernd Thu, 27 Feb 2025 17:18:05 GMT No. 25524133 >>25524162
Do a frequency analysis thingy? Not just a recording.
Bernd Thu, 27 Feb 2025 17:22:02 GMT No. 25524160
>>25524130 I was asking because if you're in an academic context, you wouldn't typically filter the noise unless explicitly asked to do so. You can apply a frequency filter quite easily in Audacity if you need to. Just play around with the values until you get something that looks good if you don't feel like calculating stuff. https://manual.audacityteam.org/man/low_pass_filter.html
Bernd Thu, 27 Feb 2025 17:22:10 GMT No. 25524162 >>25524168
>>25524133 I did fast fourier transform but I dont understand the results. Its just a collection of peaks with amplitudes decreasing exponentially, but none seem to be much larger than others. I think maybe its because my recordings are rather short? Or its harmonics? Whatever that is. I can concatenate short ones into long ones but I didnt do it yet.
Bernd Thu, 27 Feb 2025 17:23:09 GMT No. 25524166 >>25524176
>>25524127 How do I shift the phase of the noise tho?
Bernd Thu, 27 Feb 2025 17:23:24 GMT No. 25524168 >>25524200
>>25524162 >peaks with amplitudes decreasing exponentially You've got harmonics, it's completely normal. How much noise is there on the fourier transform?
Bernd Thu, 27 Feb 2025 17:24:07 GMT No. 25524170
>>25524126 I tried all kinds of grounding and its much better than with no grounding, but I still have the baseline noise with essential equipment
Bernd Thu, 27 Feb 2025 17:25:19 GMT No. 25524176 >>25524211
>>25524166 ach, sorry, what i described needs signal to be captured at 0 and 180 at the same time from the source
Bernd Thu, 27 Feb 2025 17:27:42 GMT No. 25524189
do a fourier transform and remove all the low frequencies manually
Bernd Thu, 27 Feb 2025 17:29:19 GMT No. 25524200
IMG_20250227_182617.jpg
15.19 kB, 366x1245
>>25524168 Idk, I didnt see a peak that would be obvious. I only know that its this repeating shape: Its periodic and vaguely sine-like, so its not too complex. But its close to my signal in frequency as I understand it? Since I have like 8 of it per my time window, and 2 of the signal per the same window. I also have very high frequency sine noise on all of it if you really zoom in, but it doesnt bother me much.
Bernd Thu, 27 Feb 2025 17:30:34 GMT No. 25524208
you need to view the signal in the frequency domain the pic you posted is in time domain. if it's noise from the power supply try a notch filter at 50Hz or 60Hz, whichever your cunt uses.
Bernd Thu, 27 Feb 2025 17:31:03 GMT No. 25524211 >>25524213
>>25524176 Hmmm. I dont know how to do that. It's a DC circuit with square pulses, going through a biological object.
Bernd Thu, 27 Feb 2025 17:31:34 GMT No. 25524213 >>25524216
>>25524211 What? Are you working with neurons?
Bernd Thu, 27 Feb 2025 17:31:44 GMT No. 25524216 >>25524223
>>25524213 Yes
Bernd Thu, 27 Feb 2025 17:32:06 GMT No. 25524218
he is zapping his own balls
Bernd Thu, 27 Feb 2025 17:32:07 GMT No. 25524219 >>25524258 >>25524266
ClipboardImage-1740677448.png
16.09 kB, 479x431
btw can op just do dynamic range compression in his settis? like picrel example
Bernd Thu, 27 Feb 2025 17:32:22 GMT No. 25524223 >>25524234
>>25524216 Are you sure its not the neurons then? I heard neurons can make repeating patterns
Bernd Thu, 27 Feb 2025 17:34:09 GMT No. 25524234 >>25524244
>>25524223 No, no, its very rigid noise, nothing in biological world has periodicity of exactly whatever this Hz is, across both time in recordings and individuals
Bernd Thu, 27 Feb 2025 17:36:38 GMT No. 25524244 >>25524246 >>25524277
>>25524234 If its 50 hz exactly then its most likely electrical interference from the grid.
Bernd Thu, 27 Feb 2025 17:37:08 GMT No. 25524246
>>25524244
Bernd Thu, 27 Feb 2025 17:39:30 GMT No. 25524258
>>25524219 >dynamic range compressio Hmmm interdasting. Never thought of processing sound in matlab, but its the first link in search
Bernd Thu, 27 Feb 2025 17:41:46 GMT No. 25524266 >>25524300 >>25524811
ClipboardImage-1740677943.png
5.33 kB, 222x95
>>25524219 I dont think that's the power supply or the lights.
Bernd Thu, 27 Feb 2025 17:42:32 GMT No. 25524277 >>25524284 >>25524298 >>25524319
>>25524244 The problem is I dont understand the FFT output. it gives me many many peaks and the first highest peak is at 33.33.. Hz, and then exponentially decreasing in amplitude peaks at equal intervals. The FFT tool also has like 6 modes of output, and thirty settings, so I dont know what I'm doing with it.
Bernd Thu, 27 Feb 2025 17:43:09 GMT No. 25524284 >>25524291
>>25524277 try to remove all the big frequencies and leave the normal ones alone
Bernd Thu, 27 Feb 2025 17:43:28 GMT No. 25524287
flip the power connector
Bernd Thu, 27 Feb 2025 17:43:49 GMT No. 25524291 >>25524355
>>25524284 How? High pass filter with low setting?
Bernd Thu, 27 Feb 2025 17:44:47 GMT No. 25524298 >>25524309
>>25524277 post options
Bernd Thu, 27 Feb 2025 17:44:53 GMT No. 25524300 >>25524313
>>25524266 That is ~70Hz you mean?
Bernd Thu, 27 Feb 2025 17:45:42 GMT No. 25524309
>>25524298 Hmmmm I need my cumpooter, give me a sex
Bernd Thu, 27 Feb 2025 17:45:58 GMT No. 25524313
>>25524300 if it's periodic yes
Bernd Thu, 27 Feb 2025 17:47:05 GMT No. 25524319
>>25524277 The big peaks are what you're looking to keep, it's the fundamental and it's the various harmonics. It's the strongest parts of your signal, and considering it's harmonics are at equal intervals, it's precisely the signal you're looking to keep. You need to filter out everything that is outside of those big peaks, if there's noise you should notice some smaller, insignificant peaks that need to be dealt with. Probably.
Bernd Thu, 27 Feb 2025 17:48:58 GMT No. 25524333 >>25524484
generally for an FFT you want a long sampling duration and a sampling frequency of at least 2x higher than the fastest moving signal but since this is biological it's not very fast.
Bernd Thu, 27 Feb 2025 17:51:05 GMT No. 25524355 >>25524552
>>25524291 when you see the frequency domain cant you select with your mouse a peak and delete it?
Bernd Thu, 27 Feb 2025 18:07:49 GMT No. 25524484 >>25524555
Untitled.png
47.75 kB, 1714x770
>>25524333 Ok, here's longer data and an FFT output in mode "real", and the highest peak is at "10" Theres also modes complex, magnitude, magnitude squared, phase, scaled magnitude and scaled magnitude squared for FFT
Bernd Thu, 27 Feb 2025 18:15:06 GMT No. 25524552 >>25524879
Capturhhe.PNG
95.44 kB, 1630x887
>>25524355 Its just another wave. Maybe there is such a tool but I dont know which one it might be. If its a filter, here are the filter tools. Trying high pass filter but I dont know if that looks right, dont think it does. Filtered is blue, dashed is original
Bernd Thu, 27 Feb 2025 18:15:14 GMT No. 25524555 >>25524573
>>25524484 I find that noise very strange. The sudden rise and the exponential decay suggest something switching. The alternating sign is like a digital flip-flip thingy, but viewed through some capacity. I try to make a sketch.
Bernd Thu, 27 Feb 2025 18:17:08 GMT No. 25524573 >>25524618
>>25524555 The sharp peaks is my doing, it is a square pulse going through a resistor and a capacitor (in a way). The noise is the baseline, which should be flat (flatter than this)
Bernd Thu, 27 Feb 2025 18:21:45 GMT No. 25524618 >>25524682
>>25524573 Okay, that makes it more clear. Switch it off. What's the noise then? And the FFT?
Bernd Thu, 27 Feb 2025 18:29:15 GMT No. 25524682 >>25524692 >>25524712 >>25524736 >>25524971 >>25525344
Capture1.PNG
9.95 kB, 506x401
>>25524618 This is another recording, very long, without my input, same kind of noise I think. It has largest peak at 50Hz and harmonics? at 100, 150, 200 etc. Must be power supply?
Bernd Thu, 27 Feb 2025 18:30:06 GMT No. 25524692
>>25524682 >200 well not 200 but still with a 50 hz step
Bernd Thu, 27 Feb 2025 18:31:38 GMT No. 25524712
>>25524682 So much more clear. It's not necessarily the power supply. Unless you shield everything, and properly, you'll pick up noise from all the surrounding equipment. Lamps, fans, monitors, computers, ovens, ACs, you name it.
Bernd Thu, 27 Feb 2025 18:34:55 GMT No. 25524736 >>25524762
>>25524682 ok that looks like 50hz noise and its odd harmonics (3x, 5x, etc). Try a notch filter at 50hz and 150hz since those are the biggest ones.
Bernd Thu, 27 Feb 2025 18:39:11 GMT No. 25524762 >>25524793 >>25524813 >>25525344 >>25525503
Capture2.PNG
145.99 kB, 1814x890
>>25524736 I think 50 hz works! I'm still not sure what all those settings mean or if it can be tweaked further, but the repeating shape is gone! Thanks bernds, the issue with FFT seems to be that my recordings were too short, just around one second, but I looked through the data and found a 30 seconds one, without square pulses, then FFT worked
Bernd Thu, 27 Feb 2025 18:43:43 GMT No. 25524789 >>25524868
EE is the most bernd tyre field
Bernd Thu, 27 Feb 2025 18:44:51 GMT No. 25524793 >>25524829
>>25524762 How about switching the digital signal back on? Are you still happy?
Bernd Thu, 27 Feb 2025 18:48:41 GMT No. 25524811
>>25524266 I think the 70Hz signal was your test pulse. It's always nice to include a schematic and description of the signals so there's no ambiguity.
Bernd Thu, 27 Feb 2025 18:49:06 GMT No. 25524813 >>25524856
bradley_martyn_punch.jpg
178.92 kB, 1170x1224
>>25524762 what exactly is this system and whats teh signal you are interested in? is this an audio signal? bernd is an EE can maybe help, but other bernds seem to already figured out what you need
Bernd Thu, 27 Feb 2025 18:51:13 GMT No. 25524829 >>25524843 >>25524847 >>25524926 >>25525344 >>25525630
Capture3.PNG
65.09 kB, 1375x867
>>25524793 Hmmmm its better but still kinda ass? Strange
Bernd Thu, 27 Feb 2025 18:52:48 GMT No. 25524843 >>25524879
>>25524829 what options do you have for low pass filters? the LP filter wiill smooth your signal and get rid of the sudden jolts
Bernd Thu, 27 Feb 2025 18:53:25 GMT No. 25524847 >>25524849
>>25524829 you could try averaging. white noise zeroes out and you're left with the main signal.
Bernd Thu, 27 Feb 2025 18:53:51 GMT No. 25524849 >>25524852
>>25524847 >you could try averaging averaging filters are Low-pass filters
Bernd Thu, 27 Feb 2025 18:54:18 GMT No. 25524852 >>25524857 >>25525008
>>25524849 o rly?
Bernd Thu, 27 Feb 2025 18:54:38 GMT No. 25524856 >>25524880
>>25524813 It's a DC circuit where there is an unknown complex "resistor" and unknown less complex "capacitor", and the idea is to describe those unknowns by subjecting them to whatever current of any shape you can think of
Bernd Thu, 27 Feb 2025 18:54:40 GMT No. 25524857 >>25525008
>>25524852 yes
Bernd Thu, 27 Feb 2025 18:56:02 GMT No. 25524868
>>25524789 Yes, I'm hooooofing an autism
Bernd Thu, 27 Feb 2025 18:57:21 GMT No. 25524879 >>25524897
>>25524843 The menu on the right of pic here >>25524552 Each option may or may not have sub-options
Bernd Thu, 27 Feb 2025 18:57:21 GMT No. 25524880 >>25524923 >>25524926
>>25524856 do you want to measure the resistance and capacitance? do you create the input signal digitally? a ramp signal or chirp signal is typically used to measure system functions in your case system function should consist of the resistor and capacitor in series, right?
Bernd Thu, 27 Feb 2025 18:58:57 GMT No. 25524897 >>25524940
>>25524879 a normal FIR filter should be fine, try the Low-pass option for MPR filters or look in window filters
Bernd Thu, 27 Feb 2025 19:00:38 GMT No. 25524910
FFT-Time-Frequency-View-540.png
97.51 kB, 540x378
Fast fourier transform will get the job done.
Bernd Thu, 27 Feb 2025 19:04:21 GMT No. 25524923
>>25524880 What I mostly do is that I'm amplifying the activity of a living cell, and I record that activity and responses to my actions. Electrically its (very roughly) a sphere of insulator with typical specific capacitance, described by total capacitance, and this insulator is bridged by multiple types of resistors (different rules for each type) that are behaving simultaneously, giving me some total resistance. I control the current either manually or with high frequency feedback which allows me to set voltage in the system to where I want it by quickly altering the current against intrinsic unknown currents arising in the system
Bernd Thu, 27 Feb 2025 19:04:37 GMT No. 25524924
Fourier.jpg
95.32 kB, 380x465
Give her le D.
Bernd Thu, 27 Feb 2025 19:04:40 GMT No. 25524926 >>25524933
>>25524829 >ass? >>25524880 >do you create the input signal digitally? Maybe your square wave is square only in your heda? We've all seen plenty of ringing. And the filter works fine, but you have two sources of noise?
Bernd Thu, 27 Feb 2025 19:05:19 GMT No. 25524933 >>25524937
>>25524926 >We've all seen plenty of ringing. Wuts that
Bernd Thu, 27 Feb 2025 19:06:15 GMT No. 25524937 >>25524950
>>25524933 You want the red one, but you get the yellow.
Bernd Thu, 27 Feb 2025 19:06:56 GMT No. 25524940 >>25525022
>>25524897 You mean I should apply few filters one after another?
Bernd Thu, 27 Feb 2025 19:08:00 GMT No. 25524950 >>25524981 >>25524985
>>25524937 I assume I need an oscilloscope before my ADC to see that?
Bernd Thu, 27 Feb 2025 19:08:04 GMT No. 25524954 >>25524971
Do frequency transform to figure out what you need to filter out
Bernd Thu, 27 Feb 2025 19:09:51 GMT No. 25524971
>>25524954 >>25524682 Its power supply I guess. But maybe something else too
Bernd Thu, 27 Feb 2025 19:12:01 GMT No. 25524981
anime girl thinking.jpg
497.28 kB, 3840x2160
>>25524950 I mean, after DAC. Uhhhhhhhhh
Bernd Thu, 27 Feb 2025 19:12:32 GMT No. 25524985 >>25525013
>>25524950 Yes, an oscilloscope would be nice for jobs like this. For giggles, and to test this ringing theory, you may try to decrease the drive of your digital output, it will lose the initial sharpness but it will decrease the ringing quite a bit.
Bernd Thu, 27 Feb 2025 19:15:15 GMT No. 25525008 >>25525021 >>25525229
Figure_1.png
111.85 kB, 1000x500
Screenshot 2025-02-27 201500.png
44.45 kB, 481x320
>>25524857 >>25524852 an average filter sums over multiple samples of a signal and then outputs their average. any instantaneous jiggle of the signal is supressed by the summing, which leads to LP characteristics. see pic frequency response of average filters with different sizes
Bernd Thu, 27 Feb 2025 19:16:10 GMT No. 25525013 >>25525047
>>25524985 >decrease the drive What does that mean? Sampling rate?
Bernd Thu, 27 Feb 2025 19:17:08 GMT No. 25525021 >>25525113
>>25525008 Why is it teethed like that?
Bernd Thu, 27 Feb 2025 19:17:18 GMT No. 25525022 >>25525067
>>25524940 >You mean I should apply few filters one after another? since you care about the absolute voltage and only want to supress sudden changes in the signal you will need an low-pass filter remove the other filters you dont need them please show what options you have foe low pass filters and window filters
Bernd Thu, 27 Feb 2025 19:19:42 GMT No. 25525047 >>25525092
>>25525013 Dunno. I'm but a codemonkey. Digital gates has some resistors-like things which limits the current they will output to maintain the desired voltage level. More current, faster switching. Also, more current, much more ringing. Sometimes you can set this in software on silly boards, like a raspberry pi pico. If ringing is truly your problem, you'll have to tune you circuit. Or use cleaner signals. >a ramp signal or chirp signal is typically used to measure system functions
Bernd Thu, 27 Feb 2025 19:21:55 GMT No. 25525067 >>25525132
>>25525022 Low pass window filter has 5 types to select from, Hanning, Parzen, Welch, Kaiser[1] and rectangular, end of first band, start of second band, and number of terms.
Bernd Thu, 27 Feb 2025 19:24:23 GMT No. 25525092 >>25525137
>>25525047 Well, I do need an oscilloscope I suppose, to verify the amplifier. It is supposedly a mature piece of rack tech, very niche and expensive, but you never know.
Bernd Thu, 27 Feb 2025 19:26:12 GMT No. 25525113
>>25525021 >Why is it teethed like that? a FIR average filter has a system function that has a sin component with creates the periodicity in the spectrum
H(s) = \frac{\sin (\omega N)}{\omega N} e^{-0.5 s N}
where N is the number of taps in the average filter
Bernd Thu, 27 Feb 2025 19:27:59 GMT No. 25525132 >>25525184
0f52b8109bcd16e6cc0f52fa088de287b45d6457a5946c5e352b5361bb17171d.png
99.62 kB, 1630x887
>>25525067 do a rectangular filter and show the options pls also can you select this and show options please?
Bernd Thu, 27 Feb 2025 19:28:22 GMT No. 25525137 >>25525203
>>25525092 It's not that simple. Even the best signal generator must be tuned to a signal receiver if you want to maximalize power transfer/minimalize noise. Everything goes into a billion variables differential equation, and at the end of the day, everything will depend on everything else.
Bernd Thu, 27 Feb 2025 19:33:31 GMT No. 25525184 >>25525232 >>25525294
Capture4.PNG
58.05 kB, 1370x879
Capture5.PNG
57.88 kB, 1351x865
>>25525132 Rectangular and low pass low pass seems to attack specifically the signal (peaks that go up and down, 4 in the window)
Bernd Thu, 27 Feb 2025 19:36:22 GMT No. 25525203 >>25525268
>>25525137 As of now I can only rely on oscilloscope that is part of the amplifier. It has outputs to put a separate oscilloscope in the loop but I dont have it at hand. There is hardware calibration protocol for the amplifier which I did, but I'll see if this "ringing" is an issue, it is worth checking out
Bernd Thu, 27 Feb 2025 19:39:19 GMT No. 25525229
>>25525008 ah I didn't know that. then you wouldn't want to do averaging when you're intersted in the signal close to the sampling requency.
Bernd Thu, 27 Feb 2025 19:39:30 GMT No. 25525232 >>25525293 >>25525294
>>25525184 yes, the window funciton rectangular basically does an average filtering this means the peaks in the signal will be "smeared" you can reduce the peaks futher by using different windows which have even faster declining frequency response, use more terms, or jut reduce first and second band until your output looks fine for the second one: you dont need to comput this many terms, 16 is fine reduce the end of passband and lower the stopband (try 100 and 200) error can be 3 db, but attenuation you can try reducee to 60 dB or 80 dB post the results
Bernd Thu, 27 Feb 2025 19:43:08 GMT No. 25525268 >>25525306
>>25525203 >hardware calibration protocol Tell us more. This is how riveting the rpi pico api documentation is on the subject. What do you have?
Bernd Thu, 27 Feb 2025 19:46:51 GMT No. 25525293 >>25525309 >>25525321
>>25525232 I do want to keep the peaks though, both amplitude and shape. What is between the peaks I want mostly flat
Bernd Thu, 27 Feb 2025 19:46:54 GMT No. 25525294
Window_function_and_frequency_response_-_Rectangular.png
133.60 kB, 2560x1280
Window_function_and_its_Fourier_transform_–_Hann.png
161.94 kB, 2560x1280
>>25525232 >>25525184 *other windows like kaiser or hann window do not weight all their samples to which the window is applied equally, meaning that effectively your waveform is less "smeared" also the frequency response for these windows declinnes much faster than for rectangular window (average filter) you can check the FFT to see the difference for your signal
Bernd Thu, 27 Feb 2025 19:48:51 GMT No. 25525306
>>25525268 It's all closed source, hardware calibration in my case means attaching a known value resistor and resistor/capacitor combo to the working end and then waiting 20 minutes.
Bernd Thu, 27 Feb 2025 19:49:16 GMT No. 25525309 >>25525344
>>25525293 >I do want to keep the peaks though, both amplitude and shape. What is between the peaks I want mostly flat then you want the complete opposite of what i thought you want. can you remove the filters and show the frequency spectrum/FFT again? chances are you want a band-pass filter
Bernd Thu, 27 Feb 2025 19:51:15 GMT No. 25525321 >>25525329
>>25525293 Why don't you post a pic of a noisy signal that's a problem? These are clearly flat enough already that you can easily detect the peaks.
Bernd Thu, 27 Feb 2025 19:53:08 GMT No. 25525329 >>25525334
>>25525321 He's the one creating the peaks.
Bernd Thu, 27 Feb 2025 19:54:11 GMT No. 25525334 >>25525338 >>25525362 >>25525449
>>25525329 Then I don't understand what he's trying to do. You can just create the peaks without noise.
Bernd Thu, 27 Feb 2025 19:55:19 GMT No. 25525338
>>25525334 no the noise seems to come from the system he wants to attenuate it somehow with a filter these peaks he creates himself and wants to keep
Bernd Thu, 27 Feb 2025 19:55:24 GMT No. 25525339 >>25525347 >>25525383
ClipboardImage-1740686118.png
55.75 kB, 952x186
when you're tallking about picoamps do you think that's thermal noise coming from the resistor and capacitor? that's kinda cool.
Bernd Thu, 27 Feb 2025 19:56:19 GMT No. 25525344 >>25525348 >>25525422
>>25525309 I found long duration recording with similar issue and FFT'd it here >>25524682 High-pass with notch at 50hz worked well for that recording >>25524762 but not so well for original recording >>25524829 which is weird could be all the other settings like term number or what else that I dont know the meaning. Or another noise source. Could be actual biology too, but I'd expect biological noise to have another shape and time constants. Onionbernd mentioned "ringing" and I'll test that when I steal an amplifier from someone
Bernd Thu, 27 Feb 2025 19:56:33 GMT No. 25525347
>>25525339 >talking about picoamps yes this is an insanely weak signal it has to be thermal noise
Bernd Thu, 27 Feb 2025 19:56:34 GMT No. 25525348
>>25525344 >amplifier I mean oscilloscope
Bernd Thu, 27 Feb 2025 19:57:35 GMT No. 25525362 >>25525386
>>25525334 He wants to measure something. Maybe he wants more clean samples after the peak. In his current setup he gets only a couple of clean samples after the ramp-up and then the signals quickly peters out to noise.
Bernd Thu, 27 Feb 2025 20:00:58 GMT No. 25525383 >>25525413
>>25525339 A high-resistance neuron will have resting membrane currents in order of single picoamps and even femtoamps. Its resolvable with the amplifier that I work with, it amplifies "that". If hundreds of picoamps are involved then its either not a neuron (could be an astrocyte) or it has a hull breach and it is dying rapidly.
Bernd Thu, 27 Feb 2025 20:01:18 GMT No. 25525386 >>25525449
>>25525362 Then it would still help to see an example of the problem.
Bernd Thu, 27 Feb 2025 20:04:31 GMT No. 25525413
>>25525383 oh this is more interesting than I initially thought.
Bernd Thu, 27 Feb 2025 20:05:36 GMT No. 25525422
>>25525344 could you try out different windowing functions? the notch filter works well to get rid of the frequency component at 50 Hz but this component is not the pulse you want to see? for 50 Hz it should be periodic every 20 ms
Bernd Thu, 27 Feb 2025 20:08:20 GMT No. 25525449 >>25525483 >>25525486 >>25525503 >>25525505
Untitled223.png
19.72 kB, 1714x715
>>25525386 >>25525334 If that explains it
Bernd Thu, 27 Feb 2025 20:12:27 GMT No. 25525483 >>25525570
>>25525449 since these peaks come every 50 ms this means that the frequency of you swapping the voltage has to be at 20 Hz, is this what you configured?
Bernd Thu, 27 Feb 2025 20:12:48 GMT No. 25525486 >>25525549
>>25525449 Milli-volts, and pico-amps. You'll pick a butterfly farting in straya. Which parts of the output signal is interesting to you? The delay between the the input and the output switch? The tail of the decay?
Bernd Thu, 27 Feb 2025 20:13:26 GMT No. 25525496
What a threda.
Bernd Thu, 27 Feb 2025 20:14:24 GMT No. 25525503 >>25525630
>>25525449 why did you say you were happy with the notch filter? it completely removed the peaks you wanted to see, no? >>25524762 this is confusing
Bernd Thu, 27 Feb 2025 20:14:30 GMT No. 25525505 >>25525640
>>25525449 And then you're measuring the peaks? That should be easy enough here. You said sometimes you're getting noise that's a problem: >>25524130 but you never posted an example of it.
Bernd Thu, 27 Feb 2025 20:20:23 GMT No. 25525549 >>25525574 >>25525599 >>25525640 >>25525645
Captures - Copy.PNG
53.69 kB, 1783x790
>>25525486 Going to a better example of filtering goal, above is the issue I have with the baseline, with possibly a signal in the middle. Below is filtering result using a 50hz notch filter in this threada. Signal is still there but to left and right of it seems to be newly-created artifacts.
Bernd Thu, 27 Feb 2025 20:23:40 GMT No. 25525570
>>25525483 Its a repeating pattern, 10ms holding voltage, 50ms square step to +10mV relative to holding, then back to 40ms of holding voltage. A square step. Repeated 10 times without gaps.
Bernd Thu, 27 Feb 2025 20:24:01 GMT No. 25525574
>>25525549 a notch filter wont fix this can you try different window filters and see if they fix it?
Bernd Thu, 27 Feb 2025 20:25:47 GMT No. 25525594
Does bernd know about liftering?
Bernd Thu, 27 Feb 2025 20:26:12 GMT No. 25525599
>>25525549 in this plot the signal being evened out comes from the high pass filter is this more what you are looking for? if you dont want the artifacts then make it just a high pass
Bernd Thu, 27 Feb 2025 20:30:00 GMT No. 25525630 >>25525644
>>25525503 It didnt >>25524829 they were just offset. I have two datasets in this thread, one short with peaks, about 1s, another long without peaks, about 30s. Short one did not work well with the fourier transform, because it was short. But longer one did work. They both experienced same kind of interference at the time of recording. Different sampling rate though... actually sampling rate might be relevant here, because I may be applying same filter but the tool expects same sampling rate for different recordings. The one with the peak is some weird ass number like 6.03fuck khz or something, I dont remember. That might explain why filter isnt good looking, I put 20khz for both
Bernd Thu, 27 Feb 2025 20:31:05 GMT No. 25525640
>>25525505 >>25525549
Bernd Thu, 27 Feb 2025 20:31:39 GMT No. 25525644 >>25525662
>>25525630 >Different sampling rate though yes, this is likely the issue if you fit a filter for a signal that is for one sample rate then the filter wont work for a signal with a different sample rate
Bernd Thu, 27 Feb 2025 20:31:47 GMT No. 25525645
>>25525549 This picture clarifies the ringing somewhat. It's not. The noise looks the same near and far away from the signal. I would check it anyway.
Bernd Thu, 27 Feb 2025 20:35:07 GMT No. 25525662
>>25525644 I just remembered it now. The protocol with peaks is really really legacy stuff from different era, different hardware, different country even, its around because I have a nice analysis tool written for this specific sampling rate. I really should get rid of it, or rewrite it maybe. But it also works, so maybe I sneed it
Bernd Thu, 27 Feb 2025 21:44:34 GMT No. 25526064
oekaki(1).png
9.03 kB, 500x500
After fucking around and seeing that sampling rate wasnt the core issue I feel IIR filters is where the solution is at. But its even more complex and I'm fully spent now Thx bern fun was had
Thread interest score: 7.2 Thread size: 360.71 kB